Distributed opportunistic spectrum access with dynamic users: A game-theoretic learning approach

In this paper we study the channel selection problem with dynamic users in distributed opportunistic spectrum access systems. Different with the ordinary assumption that users have data to transmit all the time, we assume that users transmit data randomly in each slot. The dynamical transmitting model makes the active number of users time-varying and channel interference levels dynamic. It makes a big difference with the static user set situation. To maximize the throughput of each user, we formulate the problem as a non-cooperative game and prove it to be a potential game in homogeneous active probability environments. The throughput of each user and total system both achieve the maximization at the same time. The stochastic learning approach has been adjusted to the dynamic users situations. The results show that stochastic learning approach algorithm converges and achieves near optimal throughput performance with a good fairness in both homogeneous and heterogeneous systems.